Product Data Scientist (Mobile Game)

We are a small remote team building a Unity mobile game. We run UA through AppLovin Axon, Meta, and Google Ads. Our current MMP is Adjust, and we plan to move to Singular. Experiments and analytics are handled in Firebase. We store data in BigQuery, use Metabase for reporting, and can connect BigQuery to Looker Studio if needed.

We’re looking for someone who enjoys digging into player behavior and making data-driven decisions. You will own hypotheses, experimentation, and applied ML (churn, early LTV, segmentation). Dashboards and the reporting layer are supported by the team, so you can focus on analysis and impact.

Help the product team understand which changes truly impact retention, churn, user acquisition quality, and monetization, and identify behavioral player segments using data mining and ML.

We prioritize validation over abstract ideation. We value your input on hypothesis feasibility, but the roadmap is primarily driven by the Product Owner. We want a strong outside perspective: help discuss options, estimate potential impact, risks, and the cost of validation, and together choose which hypothesis is most worth building and how to measure it correctly.

• Refine hypotheses into measurable statements: which metric should move, for whom, and over what time window

• Help decide what to test first by estimating expected impact, confidence, and implementation cost

• Design measurement for product changes: events, audiences, time windows, and guardrail metrics

• Evaluate impact: Firebase A/B tests, holdouts, correct interpretation of results, and control for channel and geo skews

• Deliver clear conclusions for PO and PM: what’s proven, what needs re-validation, and what to do next

• Cohort analysis and breakdowns by acquisition sources (Axon, Meta, Google), geo, devices, and versions

• Churn drivers: pre-churn patterns, friction points, and aha moments

• Segment players by play style, progression, ad sensitivity, and purchase propensity

• Churn prediction and early value estimation (early LTV)

• Clustering and segment identification with practical recommendations for product and marketing

• Association rules and relationship mining: which behavior combinations lead to retention, purchase, or churn

• Survival analysis to model time to churn and risk factors

• Pull data from BigQuery and build datasets for analysis and modeling

• Validate metrics and event quality, and help improve event tracking and metric definitions

• Automated exports to Google Sheets or Excel exports, or CSV dumps are sufficient for the start, but they must be clean and self-explanatory: clear column names, definitions, units, time period, filters, and sources.

Important: we handle visualization and BI smart reporting layer internally.